Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 1, p. 385 — regret-balancing grid forecast is ε-calibrated against adaptive Nature

Proved
AsympCalib.Main.theorem_1

by mikedeng1 · Oct 6, 2026 · Mathlib 0df444a (Lean v4.33.1)

adaptive-adversarycalibrationin-probabilityp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1randomized-forecast

Fix ε>0\varepsilon>0ε>0 and an integer grid parameter k≥1k\ge1k≥1 with k≥1/ε2k\ge1/\varepsilon^2k≥1/ε2. Let a randomized forecaster choose a grid probability i/ki/ki/k after each finite history according to a distribution satisfying the regret-flow conservation equations (2), computed from earlier rounds. Nature may use any randomized rule depending on past realized forecasts and outcomes, but does not observe the current forecast draw. If CtC_tCt​ is the realized quadratic calibration score, then there is a time t1t_1t1​ such that

Pr⁡σ,A{Ct<ε}>1−εfor every t≥t1.\Pr_{\sigma,A}\{C_t<\varepsilon\}>1-\varepsilon \qquad\text{for every }t\ge t_1.σ,APr​{Ct​<ε}>1−εfor every t≥t1​.

This identifies the particular regret-balancing algorithm as an ε\varepsilonε-calibrated forecaster against every adaptive Nature. It is stronger in content than merely asserting that some calibrated forecaster exists.

Formalization Note The paper writes k≃1/ε2k\simeq1/\varepsilon^2k≃1/ε2; the statement uses k≥1/ε2k\ge1/\varepsilon^2k≥1/ε2. Its limit condition is expressed as an eventual strict probability bound; the proof yields convergence of this probability to 111. The extra explicit k>0k>0k>0 prevents a degenerate grid. The grid has k+1k+1k+1 points, encoded as vectors (1−i/k,i/k)(1-i/k,i/k)(1−i/k,i/k), and the vectors are distinct. Round sss uses the forecaster distribution from the history prefix before sss, with Rs−1R_{s-1}Rs−1​ represented by regret over that prefix. The event concerns the realized score CtC_tCt​, not C~t\widetilde C_tCt​.

Preamble
import Mathlib
import Definitions.Def_AsympCalib_Main_Setting

noncomputable section
Formal statement
namespace AsympCalib.Main

theorem theorem_1 (ε : ℝ) (hε : 0 < ε)
    (k : ℕ) (hk0 : 0 < k) (hk : 1 / ε ^ 2 ≤ (k : ℝ))
    (σ : History → PMF (Fin (k + 1)))
    (hσ : IsConservationForecaster k σ)
    (A : History → PMF (Fin 2)) :
    ∃ t₁ : ℕ, ∀ t ≥ t₁,
      ENNReal.ofReal (1 - ε) <
        (CalibratedCE.Forecast.histLaw (forecaster k σ) A t).toOuterMeasure
          {h : History | CalibratedCE.Forecast.calibScore2Hj h 1 < ε} := by sorry

end AsympCalib.Main
Source
Foster and Vohra, Asymptotic calibration, Biometrika 85 (1998), p. 385, Theorem 1; p. 383, Definition (ε-calibration); p. 388, Corollary (§6); https://doi.org/10.1093/biomet/85.2.379
Human review
  • Endorsed by Shuze Chen · Oct 6, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 6, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me